Papers by Douglas Eadline
Making Prolog parallel
AI Expert archive, Jul 1, 1989
Implementing Prolog on Distributed Systems
: A method for distributing Prolog across multiple processors is described. Design criterion and ... more : A method for distributing Prolog across multiple processors is described. Design criterion and rationale are delineated along with the assumptions behind the methodology. The underlying assumptions are based on the need to provide each processor with maximum sequential work. The parallelization mechanism was implemented using a meta-interpreter that includes new predicates for dependency detection, job queuing, queue solving, and queue maintenance. Debugging and side effects issues are addressed. Benchmark results are also presented. 1. History In 1988, we began the design of Prolog that would operate in parallel over any number of computing processors. As a commercial company, our goal was to provide a high level tool for systems programmers using parallel or distributed systems. To date, the result of our efforts have produced several prototypes and a commercial version which goes by the name of n-parallel PROLOG. During the course of development, we learned several things about ...
Maximizing "Return on Investment": Prolog Extensions for Effective Decision Support Applications
: Extensions to Prolog that offer the user portable Graphical User Interfaces (GUIs), transparent... more : Extensions to Prolog that offer the user portable Graphical User Interfaces (GUIs), transparent access to foreign databases, and scalable performance are presented. These extensions have been implemented on top of standard Prolog and can be shown to provide a powerful level of abstraction that ultimately leads to reduced costs and better return on investment for Decision Support (DS) applications. Introduction Building enterprise wide decision support (DS) applications has pushed procedural languages (i.e. C/C++) to the limit. Building such applications with traditional tools can lead to large costs and large delays in application development. Advances in object technology address some of these areas, but do not offer a complete solution. Building complex and sophisticated DS applications must be driven by cost savings and provide an acceptable return on investment. The ability of traditional tools and methods to meet these needs is highly questionable. Fortunately, Prolog has been ...
Prolog/Database Parallel Programming: The '?-a(X),b(Y),c(X,Y)' Challenge
: Two methods of multiple/large/foreign databases processing using massively parallel computers a... more : Two methods of multiple/large/foreign databases processing using massively parallel computers are described. The first method employs a Distributed Cache Subsystem (DCS) offering the total memory of parallel processing nodes as a very large read-only disk cache for other parallel processing nodes. The second method is based on a Database Slicing (DS) algorithm that provides logical partitioning of the entire database between the processing nodes. These methods were implemented using n-parallel PROLOG from Paralogic. Results of experiments on an nCUBE-2 parallel computer are presented. INTRODUCTION It is well-known that Prolog language can be used efficiently as a 4GL query language to access and process relational databases [1]. Furthermore the Prolog language is well suited to run on a massively parallel computers. In recent years a large amount of work has been done investigating parallel Prolog execution models [2]. Our interest is determining whether parallel Prolog running on ...
Design and Implementation of a Prolog-to-C Compiler
A WAM-based Prolog-To-C (PTC) compiler is described. It is written in the C language and used tog... more A WAM-based Prolog-To-C (PTC) compiler is described. It is written in the C language and used together with Paralogic's Prolog interpreter. Prolog programs may consist of several independently compiled modules. The principles of modularization are described. Mixed programming features (Prolog-C and CProlog programming interfaces) are also explained. The PTC compiler is portable and has been ported to several different hardware platforms (PC, Sun, SGI, HP, nCube-2, Transputer.). Some statistical data on the compilers efficiency are shown.
DUMmIES SUN AND AMD SPECIAL EDITION High Performance Computing For Dummies®, Sun and AMD Special Edition
• Pick out hardware and software • Find the best vendor to work with • Get your people up to spee... more • Pick out hardware and software • Find the best vendor to work with • Get your people up to speed on HPC Learn to:
Mapping Search Graphs Onto Arbitrary Processor Networks (or Making PROLOG Parallel)
Logic programming and particularly PROLOG have long been considered good candidates for parallel ... more Logic programming and particularly PROLOG have long been considered good candidates for parallel execution. The absence of any formal control structures have prompted such claims. There are, however, some practical concerns when actually implementing a parallel version of standard PROLOG. Specifically, an economical balance between identifying parallel tasks and solving parallel tasks must be obtained. Furthermore, the balancing mechanism should operate at run time and have a low overhead. Paralogic, Inc. has implemented a parallel version of PROLOG on the INMOS transputer. The current release supports the "standard" Edinburgh syntax and will run PROLOG programs on one to "n" transputers without modification. 1. PROLOG as a Parallel Language If implemented properly, parallel PROLOG would have several advantages over conventional languages. First, parallel PROLOG will be independent of the number and connection of parallel processors. Programs would be expected to run on one to "n" processors without modification. Second, parallel PROLOG will implicitly determine parallelism. Though ultimate parallel performance will depend upon the parallel algorithm, parallel PROLOG will find enough parallelism to keep a number of processors busy doing useful work. Third, parallel PROLOG will be portable at the source code level between sequential computers and any parallel computer. Standard PROLOG programs that have been written on sequential machines will run virtually unmodified on parallel computers. By design, PROLOG has no formal control structures. This feature is both a blessing and a curse for the PROLOG language. The lack of control structures is due to the resolution mechanism used by PROLOG. The software engineer programs in PROLOG by declaring a database of facts and rules. The program is "run" by presenting the database with a question or query. Program execution terminates when all solutions to the query have been found. The resolution mechanism takes a query and derives an answer from the database the same way every time. In short, all pure PROLOG programs function the same. Because the resolution mechanism is the same for all PROLOG programs, parallelizing PROLOG can be accomplished by parallelizing the resolution mechanism. Mapping Search Graphs Onto Arbitrary Processor Networks ________________________________________________________________ ___________________________ There should be no consequence to the PROLOG database (source code) and any PROLOG program should be transportable between sequential and parallel hardware platforms. Although research has provided resolution mechanisms [1-7], there are several practical concerns outside the scope of the resolution mechanism that must be addressed when parallelizing PROLOG. The absence of control structures is, to some, the bane of PROLOG. Although virtually any type of program can be written in PROLOG, many programmers have problems using PROLOG because it does not fit into the "control structure paradigm". When used for the right type of problem, however, PROLOG has been shown to be more efficient then conventional languages [8]. 2. Search Tree Representations of PROLOG Programs Implementing PROLOG in parallel requires the mapping of search trees onto an arbitrary network of processors or transputers. In PROLOG, a search tree is a graphical representation of the possible paths a PROLOG program may take while executing. The resolution mechanism used by PROLOG will continue along one path until an answer is found or failure occurs. In either case, the resolution mechanism does not stop but, continues to search other paths for different answers. This process is referred to as "backtracking" in sequential PROLOG. In parallel PROLOG, solutions may come from one path or another and is appropriately referred to OR parallelism. The listing below is a simple multiple solution PROLOG program. For more information on PROLOG see reference [9]. Essentially there are twelve facts and three rules. The facts have to do with the relationships within a family. For instance, the first fact states that "dave is the father of steve", the second states that "dave is the father of doug", and so on. The first two rules define the "parent_of" relation. For instance, the first rule states that "a variable X is the parent of Y if X is the father of Y". Because parents can be either mothers or fathers, we need a second rule that states "a variable X is the parent of Y if X is the mother of Y".
Abstract: A general law is proposed that states "a large numbers of slower processors may be... more Abstract: A general law is proposed that states "a large numbers of slower processors may be better than a small number of faster processors for I/O-dominant applications". The need for such a guideline is demonstrated because simple linear sums of individual processor performances do not provide an accurate estimation of I/O performance for a parallel computer. Furthermore, the law was formulated to allow better cost estimations when choosing the number and type processor for a Massively Parallel (MP) I/O application. The law is confirmed with a simple proof, analytical model, simulation, and benchmarking. A Distributed Cache Subsystem (DCS) technique is proposed to further improve the performance of the MP computers running I/O-dominant applications. Using simulations and benchmarks the DCS technique has shown the potential to achieve very high performance using standard sequential file systems. The general conclusion proposed is that a MP computer containing a small num...
Implementing Prolog on Distributed Systems: N-parallel Prolog
A method for distributing Prolog across multiple processors is described. Design criterion and ra... more A method for distributing Prolog across multiple processors is described. Design criterion and rationale are delineated along with the assumptions behind the methodology. The underlying assumptions are based on the need to provide each processor with maximum sequential work. The parallelization mechanism was implemented using a meta-interpreter that includes new predicates for dependency detection, job queuing, queue solving, and queue maintenance. Debugging and side effects issues are addressed. Benchmark results are also presented.
A general law is proposed that states "a large numbers of slower processors may be better th... more A general law is proposed that states "a large numbers of slower processors may be better than a small number of faster processors for I/O-dominant applications". The need for such a guideline is demonstrated because simple linear sums of individual processor performances do not provide an accurate estimation of I/O performance for a parallel computer. Furthermore, the law was formulated to allow better cost estimations when choosing the number and type processor for a Massively Parallel (MP) I/O application. The law is confirmed with a simple proof, analytical model, simulation, and benchmarking. A Distributed Cache Subsystem (DCS) technique is proposed to further improve the performance of the MP computers running I/O-dominant applications. Using simulations and benchmarks the DCS technique has shown the potential to achieve very high performance using standard sequential file systems. The general conclusion proposed is that a MP computer containing a small number of hig...
Redefining Scalable OpenMP and MPI Price-to-Performance with Numascale ’ s
The NumaConnect cluster excels at both OpenMP and MPI computing within the same shared memory env... more The NumaConnect cluster excels at both OpenMP and MPI computing within the same shared memory environment. No extra software or program modifications are needed to take advantage of the entire system. Results for the NASA Advanced Supercomputing (NAS) Parallel Benchmarks have set a new record for OpenMP core count and problem size. OpenMP results show good scalability, with best results coming from larger problem sizes.
Design and Implementation of a Prolog-to-C Compiler
A WAM-based Prolog-To-C (PTC) compiler is described. It is written in the C language and used tog... more A WAM-based Prolog-To-C (PTC) compiler is described. It is written in the C language and used together with Paralogic's Prolog interpreter. Prolog programs may consist of several independently compiled modules. The principles of modularization are described. Mixed programming features (Prolog-C and CProlog programming interfaces) are also explained. The PTC compiler is portable and has been ported to several different hardware platforms (PC, Sun, SGI, HP, nCube-2, Transputer.). Some statistical data on the compilers efficiency are shown.
Practical Data Science With Hadoop And Spark
Maximizing " Return on Investment " : Prolog Extensions for Effective Decision Support Applications
Extensions to Prolog that offer the user portable Graphical User Interfaces (GUIs), transparent a... more Extensions to Prolog that offer the user portable Graphical User Interfaces (GUIs), transparent access to foreign databases, and scalable performance are presented. These extensions have been implemented on top of standard Prolog and can be shown to provide a powerful level of abstraction that ultimately leads to reduced costs and better return on investment for Decision Support (DS) applications.
Hadoop 2 Quick-Start Guide: Learn the Essentials of Big Data Computing in the Apache Hadoop 2 Ecosystem
Get Started Fast with Apache Hadoop 2, YARN, and Todays Hadoop Ecosystem With Hadoop 2.x and YARN... more Get Started Fast with Apache Hadoop 2, YARN, and Todays Hadoop Ecosystem With Hadoop 2.x and YARN, Hadoop moves beyond MapReduce to become practical for virtually any type of data processing. Hadoop 2.x and the Data Lake concept represent a radical shift away from conventional approaches to data usage and storage. Hadoop 2.x installations offer unmatched scalability and breakthrough extensibility that supports new and existing Big Data analytics processing methods and models. Hadoop 2 Quick-Start Guide is the first easy, accessible guide to Apache Hadoop 2.x, YARN, and the modern Hadoop ecosystem. Building on his unsurpassed experience teaching Hadoop and Big Data, author Douglas Eadline covers all the basics you need to know to install and use Hadoop 2 on personal computers or servers, and to navigate the powerful technologies that complement it. Eadline concisely introduces and explains every key Hadoop 2 concept, tool, and service, illustrating each with a simple beginning-to-end...
Implementing Prolog on Distributed Systems: N-parallel Prolog
Ilps Islp Naclp Slp, 1994
Implementing Prolog on Distributed Systems
Design and Implementation of a Prolog-to-C Compiler
Run-time system having nodes for identifying parallel tasks in a logic program and searching for available nodes to execute the parallel tasks
Experimental errors and artifacts for a time-domain lumped capacitor dielectric spectrometer
Review of Scientific Instruments, 1986
Uploads
Papers by Douglas Eadline